128 FPGA Cluster
The next generation of performance-optimized reconfigurable computing
Using 16 to 128 Spartan–6 LX150 FPGAs in a processing-optimized configuration, RIVYERA S6-LX150 represents a new generation of application specific computing and provides high-density supercomputing resources to a wide audience. Additional local RAM offers sufficient memory for all types of computation and data. From assembly of bioinformatics data to z-buffering in 3D graphics, the RIVYERA is the highest performant computer in the market.
- 16 to 128 Xilinx Spartan-6 LX150 FPGAs per machine
- scalability up to 2560 FPGAs per rack; mixed FPGA setups supported
- up to 64 GB RAM and 4 TB SDHC total
- Up to 15000 CPU cores performance (application specific)
Using 128 Xilinx Spartan–6 LX150 FPGAs in its standard configuration, one RIVYERA provides all the performance needed for extreme but cost-efficient speed-ups of your applications or even new types of analysis and competitive advantage. Directly attached 512 MByte of DRAM and the possibility to add an additional 32 GByte of micro-SD card based memory, per FPGA, and two 1 Terabyte SATA-II hard disks offer lots of memory even for memory intensive applications.
Each RIVYERA has an integrated standard off–the–shelf PC internally connected through up to seven 4-lane PCI-Express based i/o cards. A set-up as a FPGA cluster with several RIVYERA can be achieved with virtually no performance loss by using SciEngines UPLINK technology. Up to 10 RIVYERA (1280 FPGAs) can be fit into a standard rack and be connected to provide - for specific algorithms - performance comparable to the world's top 50 supercomputers.
- Computational Performance: Using up to 128 Xilinx Spartan–6 LX150 FPGAs in a fully scalable setup, RIVYERA provides an optimal cost-benefit ratio.
- Memory: Providing 512 MB DRAM for each of the 128 FPGAs, a standard RIVYERA offers enough distributed memory for any application. Of course, the regular extremely high speed block RAM of the FPGAs is also available.
- Additional Memory: Oftentimes, algorithms require look-ups in larger tables. With RIVYERA, you have the option to have one 32-GB SDHC card expansion per FPGA (4 TB total) to further enhance the distributed computation capabilities.
- x86 Processing Power: Joining standard x86 based architectures with high performance FPGA technology, RIVYERA allows your CPU to do what it has been originally designed for: Lower-performance, general purpose computing. Focusing on a hybrid computing approach, users can leave significant parts of their codes with little to no adjustments on the host-PC and only execute the most intensive parts of the application on the FPGAs.
- FPGA Acceleration: Xilinx is one of the world's leaders in high density FPGA technology and this shows in SciEngines massively parallel setups. RIVYERA S6-LX150 provides typical speedups from 1500 to 15000 CPU cores (depending on the algoritm) and thus easily surpasses competitive technologies like GPU computing or other FPGA-based products.
- High Speed Bus System: Offering interconnection technics from PCI-Express to Infiniband, SciEngines RIVYERA based platforms ensure to transport their backplane bandwidth into your IT-infrastructure.
- Unix Stability: RIVYERA utilizes an enterprise scale open source based LINUX OS. Oracle Enterprise Linux provides the reliability and power you expect from any PC operating system and seamlessly fits into your cluster set-up. It behaves "just like any other server in the network" and is simply orders of magnitude faster.
- Windows Flexebility: You would like to have Microsoft Windows preinstalled on your machine: No problem. SciEngines is Microsoft certified partner.
- Platform Support: Featuring APIs based on C, C++, Java, VHDL and Verilog, RIVYERA ensures to support nearly every platform. An easy to use communication framework together with a set of demo application enables rapid target application and algorithm integration.
Memory, Bandwidth and API
The new RIVYERA S6-LX150 has a number of configuration options and, in addition to the FPGAs block RAM, provides 512 MB DDR3 RAM and optionally up to 32 GB micro SD-card based memory per FPGA. Each FPGA accesses its local RAM with a bandwidth of up to 3.2 GB/s, which leads to an enormous data-access rate of 200+ GB/s in a 16 cards / 128 FPGA setup - if the algorithms are suitable for such distributed processing.
Addressing the FPGAs is made easy by the API's usual and well-liked features like broadcasting, individual addressing of FPGAs and (partial) reconfiguration during run-time.For elaborate documentation of SciEngines' API and how it may support your development projects, please see our documentation section.
Quality and Reliability
Our goal is to provide our customers the maximum of quality and reliabilty they could expect for their money. Using industry and military proven components and technics, SciEngines products and especially RIVYERA offers one of the most reliable high performance reconfigurable platforms worldwide. All our products are testet in an 48h stress-test to ensure minimum failures. Continuously thermal monitoring and redundant supplies allow our architecture to ensure operation even in tough environmental conditions. Some of our systems are running 24/7 at a German governmental association without any error nor fail or need of any maintenance in 5 years of operation.
|In the box
- Rack mounting hardware
- Power cords
- I/O cable (depends on option)
- Printed and electronic documentation
- Drivers and Utilities CD–ROM
- 16 to 128 Xilinx Spartan–6 LX150
- optional: 32 GB SDHC Card per FPGA
- Supermicro off-the-shelf server grade mainboard
- Xeon E3-1220 or better
- 12 GB RAM, ECC DDR3
- SATAII RAID-Controller
- up to 6 PCIe slots for i/o cards
- (standard configuration, b2o available)
- optional: up to 2–TB SATA–II harddisks
- 16 identical slots available for SciEngines FPGA cards
- dual 10⁄100⁄1000BASE–T (Gigabit) RJ–45 Ethernet interface (supporting jumbo frames)
- optional: 10G-Ethernet
- optional: fibre channel
- optional: firewire 400
- optional: firewire 800
- offered interface depends on configuration
- Fits EIA–310–D–compliant, industry–standard four–post racks and cabinets: 19 inches wide, 24 to 36 inches deep front–to–back cooling for rack enclosure
- front–to–back cooling for rack enclosure
- Support for square–hole racks or threaded racks based on mounting kit selection
- Output power: 1400W (redundant supply)
- Optional: 3kW power supply available
- Line voltage: universal input (100V to 240V AC), power factor corrected
- Maximum input current: 32.0A (100V to 120V) or 16.0A (200V to 240V)
- Frequency: autom. shifting 47Hz to 63Hz, single phase
- Specification: 2 U Sure Star 2x 650W EPS mini redundant
- PSU Reliability: MTBF: 100.000 hours at 25°C
- Certifications: FCC Part 15, Subpart B, Class B, EN55022 CISPR22 Class B, UL60950, CAN/CSA22.2 No.60950, TUV(IEC60950), CE, CB, ISO 9001, ROHS 2002/95/EC
- Description: Suitable for high available 24/ 7 continous operation request. Hot swap changeable during current running. An undisturbed service is guaranteed. Separated power input.
- 4x 80mm hot-swap fans + 2x 80mm ball bearing fan at rear
- Operating temperature:
10°C to 35°C (50°F to 95°F)
- Storage temperature:
-40°C to 70°C (-40°F to 158°F)
- Operating Humidity (non condensing):
8% to 70% RH
- Storage Humidity (non condensing):
5% to 95% RH
- Optional: Military Quality Hardware
Size and weight
- Height: 17.8cm (7 inches)
- Width: 43.7 cm (17.2 inches) for mounting in standard 19–inch rack
- Depth: 66.9 cm (27.5 inches) - please allow for additional room in the back to ensure airflow
- Weight: depends on configuration, e.g. 58.0 pounds (26.0 kg) base configuration
- Communication Test
- Communication Framework
- Target Application Specific Software Packages (optional)